Re: [Talk-GB] too many universities in Cambridge

[Lester Caine]

On 22/05/15 15:00, Dan S wrote:
>>> Andy, the operator tags are all the same, not the building names.
>> >
>> > No, they really aren't.
>> >
>> > http://www.openstreetmap.org/way/148247775 - "Churchill College
>> > (University of Cambridge)"
>> > http://www.openstreetmap.org/way/12861651 - "University of Cambridge"
>> > http://www.openstreetmap.org/way/98523431 - "Clare College (University
>> > of Cambridge)"
> but David's using a constrained and documented set of operator tags,
> not free-text. I think that's good. The oxbridge college system makes
> this a difficult case study! I think the operator tags here are
> appropriate.

The only problem with this type of data is that there is no
'cross-reference' ... my places hierarchy would have 'University of
Cambridge' as a main place, and each of the collages would then have a
place referencing 'University of Cambridge'.  One can then flag a single
'University of Cambridge' icon at a macro level, and switch to icons for
the collages at a suitable scale. That data in the map can then be
cross-referenced from the master objects ... and be highlighted when a
particular place is selected.

-- 
Lester Caine - G8HFL
-
Contact - http://lsces.co.uk/wiki/?page=contact
L.S.Caine Electronic Services - http://lsces.co.uk
EnquirySolve - http://enquirysolve.com/
Model Engineers Digital Workshop - http://medw.co.uk
Rainbow Digital Media - http://rainbowdigitalmedia.co.uk

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] too many universities in Cambridge

[Dan S]

2015-05-22 14:49 GMT+01:00 Andy Allan :
> On 22 May 2015 at 14:27, David Earl  wrote:
>> Andy, the operator tags are all the same, not the building names.
>
> No, they really aren't.
>
> http://www.openstreetmap.org/way/148247775 - "Churchill College
> (University of Cambridge)"
> http://www.openstreetmap.org/way/12861651 - "University of Cambridge"
> http://www.openstreetmap.org/way/98523431 - "Clare College (University
> of Cambridge)"

but David's using a constrained and documented set of operator tags,
not free-text. I think that's good. The oxbridge college system makes
this a difficult case study! I think the operator tags here are
appropriate.

Dan

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] too many universities in Cambridge

[David Earl]

Yes, the operator tags are the same when it is the same institution - the
colleges are independent institutions, part of the larger federation. This
is part of the complexity of this.

I'm not arguing I don't want to change anything, just that there's too much
gratuitous change which breaks real, existing products because of
hypothetical futures.The wiki analogy is wrong here I think - that's the
content. It's much more an API, as I think you were essentially agreeing,
and people go to great lengths to try to maintain backward compatibility,
only deprecating things when they absolutely have to.

And it's not so much me not wanting to change things, of course change
happens, it's random, arbitrary, incompatible change that is such a problem
to deal with. Dan's not arguing for that, and I've already said I'll look
at it and see what's involved. But not today!



On Fri, 22 May 2015 at 14:49 Andy Allan  wrote:

> On 22 May 2015 at 14:27, David Earl  wrote:
> > Andy, the operator tags are all the same, not the building names.
>
> No, they really aren't.
>
> http://www.openstreetmap.org/way/148247775 - "Churchill College
> (University of Cambridge)"
> http://www.openstreetmap.org/way/12861651 - "University of Cambridge"
> http://www.openstreetmap.org/way/98523431 - "Clare College (University
> of Cambridge)"
>
> > But also the assertion "within a few dozen miles" is wrong, as for
> > Nottingham in China.
>
> Read what I said, please:
>
> >> If there were two objects tagged as universities with
> >> identical names within a few dozen miles, I could make a guess they
> >> are the same university and write some rendering rules to suit.
>
> I make no assertion that all parts of the same university are within a
> dozen miles.
>
> I hope you realise that your tagging (using tags that imply 1200
> different universities) is causing problems, and think "what could I
> do to help other people" rather than "I don't want to change
> anything".
>
> Thanks,
> Andy
>
___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] too many universities in Cambridge

[Stuart Reynolds]

Without wanting to get into specific tags, or indeed into specific renderers, 
let’s step back and see if what we have got is what we want?
The answer is probably no, IMHO.

Take Newnham College 
(http://www.openstreetmap.org/#map=17/52.19959/0.10973&layers=H) which I know 
fairly well. It is one site - Newnham College - and is part of the university. 
But the individual buildings within it are just that - buildings. In fact, the 
distinction between the various buildings is really only of relevance to the 
people within it, given that you can walk between the extremities Strachey in 
the east to Peile in the west without every going outside, and the fact that 
much of it is student’s rooms.

So:

- we need to be able to identify the site as Newnham College
- we need to be able to identify Newnham College as part of the University of 
Cambridge
- we need to be able to name and identify the buildings on the site, and to 
have them linked to Newnham College. But we do NOT need to reference them as 
universities in their own right

So long as we use tagging and/or relationships which maintain those 
associations, we have clarity on the data and renderers can choose what to do 
with it.

Stuart

On 22 May 2015, at 14:22, Andy Allan 
mailto:gravityst...@gmail.com>> wrote:

On 22 May 2015 at 14:03, Christopher Baines 
mailto:m...@cbaines.net>> wrote:
On 21/05/15 22:39, Dan S wrote:
I don't relish bringing this up since it's a bit of a tangle, but I
noticed Cambridge has a lot more universities than I thought!
Apparently 1219, judging from the number of amenity=university tagged
objects. In real life I'm aware of two: Cambridge Uni, Anglia Ruskin
Uni.

I think that it is a poor assumption to make that there exists a one to
one mapping between objects (nodes, ways, relations) tagged with
amenity=university, and actual organisations.

Sure, but then you need to look at what is actually being tagged.
We've already heard that there are 1219 different universities in
Cambridge, so I was intrigued as to what they are. After all, I would
expect "amenity=university; name=University of Somewheresville" to be
a university. If there were two objects tagged as universities with
identical names within a few dozen miles, I could make a guess they
are the same university and write some rendering rules to suit.

But they are all different. There's a university named "Music Centre".
There's another university called "Pavillion D". There's a third
university called "Forbes Mellon Library" which is a surprising thing
to call a university. There's a bunch of little unamed universities.
And they all have different operator tags too.

I suspect these are the names of buildings, not universities. I
suspect they are operated by different sections of the one university,
but there's no easy way to tell from the operator tag without a
natural-language parser coupled with a wikipedia-based explanation of
the constituent college system.

Have a look at the data, and you'll see it's not as straightforward as
you think. Sure, there's no one-to-one mapping between the real world
and OSM features. But that's not what we're talking about here.

Thanks,
Andy

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] too many universities in Cambridge

[Andy Allan]

On 22 May 2015 at 14:27, David Earl  wrote:
> Andy, the operator tags are all the same, not the building names.

No, they really aren't.

http://www.openstreetmap.org/way/148247775 - "Churchill College
(University of Cambridge)"
http://www.openstreetmap.org/way/12861651 - "University of Cambridge"
http://www.openstreetmap.org/way/98523431 - "Clare College (University
of Cambridge)"

> But also the assertion "within a few dozen miles" is wrong, as for
> Nottingham in China.

Read what I said, please:

>> If there were two objects tagged as universities with
>> identical names within a few dozen miles, I could make a guess they
>> are the same university and write some rendering rules to suit.

I make no assertion that all parts of the same university are within a
dozen miles.

I hope you realise that your tagging (using tags that imply 1200
different universities) is causing problems, and think "what could I
do to help other people" rather than "I don't want to change
anything".

Thanks,
Andy

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] too many universities in Cambridge

[Andy Allan]

On 22 May 2015 at 11:54, David Earl  wrote:

> 2. What is a University anyway?

I'll not explore the concept of a university too far, since very
little about groups of people is relevant to OSM!

However, if you were to say "What is the physical aspect of a
university" then I would say it's a collection of one or more places -
usually parcels of ground, often with buildings. In OpenStreetMap we
tend to represent collections of entities with a relation, unless the
tagging of each part is substantially the same (i.e. no need for a
relation when you split a road and add bridge tags.

I note that the amenity=university tag is used mostly (exclusively?)
on buildings, and I think that this is incorrect. I would expect a way
tagged amenity=university to indicate that everything within that way
- buildings, gardens, carparks etc - was part of the university. So we
can probably retag things to cut down the numbers by using perimeters
around particular areas.

> 4. Constantly changing tags creates a moving target that is extremely hard
> to maintain for data consumers, and is a major off-putting factor in using
> OSM, especially if you can't manage the process because things just change
> under your feet.

Yep, I certainly agree with you there. But when things are tagged
'incorrectly' (fsvo incorrect, of course), then we need to change the
tagging.

> My view is that tags are merely
> tokens and too much is read into the words.

Yep, I've gone on at length about this with people wishing to change
the order of the characters within a particular tag - it's
infuriating.

>  bear in mind this has a
> direct financial cost to me as a freelancer supporting the University map,
> and that the University has been a big benefactor for OSM, even though they
> get the rest of the map back in return, so you really don't want to give
> them a slap in the face for doing so.

Be careful. To say that we need to support your old tagging scheme
indefinitely would seem to be a slap in the face for all our
volunteers, now and in the future.

Thanks,
Andy

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] too many universities in Cambridge

[Shaun McDonald]

> On 22 May 2015, at 12:30, David Earl  wrote:
> 
> Does the main OSM rendering understand building=university?
> 

Yes, for example this building at Heriot-Watt University: 
http://www.openstreetmap.org/way/22881764 
 where I used the 
university=building tag over 6 years ago. With the grounds of the university 
being tagged amenity=university: http://www.openstreetmap.org/way/4388535 
 I apply a similar principle to 
schools.

Mapnik will render any building value, except building=no. There are some 
building values which get special treatment.
https://github.com/gravitystorm/openstreetmap-carto/blob/master/project.mml#L468
 

https://github.com/gravitystorm/openstreetmap-carto/blob/master/buildings.mss 
 

Shaun


___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] too many universities in Cambridge

[Dan S]

2015-05-22 14:03 GMT+01:00 SK53 :
> For what it is worth, the universities in Nottingham are mapped exactly the
> same way as Cambridge

It doesn't look like that to me. A query on Nottingham finds 16
objects tagged amenity=university, of which 4 are buildings. It looks
to me like Nottingham broadly uses amenity=university at "site level"
rather than "building level" (with some exceptions).

> Finally, my general view is that the only sensible tag for consolidating
> separate campuses is to add an operator=* tag. We could do with some
> agreement about how to label campuses. Do we use "University of Nottingham
> Jubilee Campus" or "Jubilee Campus" etc. And of course some checking on the
> use of the operator tag itself.

Sounds good.

> PS. The use of an icon on the HOT map layer doesn't work for me in 1st world
> countries: but I dont regard that as terribly important, surely the HOT
> layer needs to be directed at places which most need the rendering scheme
> chosen

Oh, absolutely. Sorry for being unclear there: the HOT rendering is
just a real-life example of the rendering problem I wanted to
illustrate.

Dan

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] too many universities in Cambridge

[David Earl]

Andy, the operator tags are all the same, not the building names.

But also the assertion "within a few dozen miles" is wrong, as for
Nottingham in China.

On Fri, 22 May 2015 at 14:23 Andy Allan  wrote:

> On 22 May 2015 at 14:03, Christopher Baines  wrote:
> > On 21/05/15 22:39, Dan S wrote:
> >> I don't relish bringing this up since it's a bit of a tangle, but I
> >> noticed Cambridge has a lot more universities than I thought!
> >> Apparently 1219, judging from the number of amenity=university tagged
> >> objects. In real life I'm aware of two: Cambridge Uni, Anglia Ruskin
> >> Uni.
> >
> > I think that it is a poor assumption to make that there exists a one to
> > one mapping between objects (nodes, ways, relations) tagged with
> > amenity=university, and actual organisations.
>
> Sure, but then you need to look at what is actually being tagged.
> We've already heard that there are 1219 different universities in
> Cambridge, so I was intrigued as to what they are. After all, I would
> expect "amenity=university; name=University of Somewheresville" to be
> a university. If there were two objects tagged as universities with
> identical names within a few dozen miles, I could make a guess they
> are the same university and write some rendering rules to suit.
>
> But they are all different. There's a university named "Music Centre".
> There's another university called "Pavillion D". There's a third
> university called "Forbes Mellon Library" which is a surprising thing
> to call a university. There's a bunch of little unamed universities.
> And they all have different operator tags too.
>
> I suspect these are the names of buildings, not universities. I
> suspect they are operated by different sections of the one university,
> but there's no easy way to tell from the operator tag without a
> natural-language parser coupled with a wikipedia-based explanation of
> the constituent college system.
>
> Have a look at the data, and you'll see it's not as straightforward as
> you think. Sure, there's no one-to-one mapping between the real world
> and OSM features. But that's not what we're talking about here.
>
> Thanks,
> Andy
>
> ___
> Talk-GB mailing list
> Talk-GB@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-gb
>
___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] too many universities in Cambridge

[Andy Allan]

On 22 May 2015 at 14:03, Christopher Baines  wrote:
> On 21/05/15 22:39, Dan S wrote:
>> I don't relish bringing this up since it's a bit of a tangle, but I
>> noticed Cambridge has a lot more universities than I thought!
>> Apparently 1219, judging from the number of amenity=university tagged
>> objects. In real life I'm aware of two: Cambridge Uni, Anglia Ruskin
>> Uni.
>
> I think that it is a poor assumption to make that there exists a one to
> one mapping between objects (nodes, ways, relations) tagged with
> amenity=university, and actual organisations.

Sure, but then you need to look at what is actually being tagged.
We've already heard that there are 1219 different universities in
Cambridge, so I was intrigued as to what they are. After all, I would
expect "amenity=university; name=University of Somewheresville" to be
a university. If there were two objects tagged as universities with
identical names within a few dozen miles, I could make a guess they
are the same university and write some rendering rules to suit.

But they are all different. There's a university named "Music Centre".
There's another university called "Pavillion D". There's a third
university called "Forbes Mellon Library" which is a surprising thing
to call a university. There's a bunch of little unamed universities.
And they all have different operator tags too.

I suspect these are the names of buildings, not universities. I
suspect they are operated by different sections of the one university,
but there's no easy way to tell from the operator tag without a
natural-language parser coupled with a wikipedia-based explanation of
the constituent college system.

Have a look at the data, and you'll see it's not as straightforward as
you think. Sure, there's no one-to-one mapping between the real world
and OSM features. But that's not what we're talking about here.

Thanks,
Andy

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] too many universities in Cambridge

[Christopher Baines]

On 21/05/15 22:39, Dan S wrote:
> I don't relish bringing this up since it's a bit of a tangle, but I
> noticed Cambridge has a lot more universities than I thought!
> Apparently 1219, judging from the number of amenity=university tagged
> objects. In real life I'm aware of two: Cambridge Uni, Anglia Ruskin
> Uni.

I think that it is a poor assumption to make that there exists a one to
one mapping between objects (nodes, ways, relations) tagged with
amenity=university, and actual organisations.

For example, you would not count the roads in a town by looking at all
ways tagged with highway=(primary|secondary) in that town, as more than
one way can make up just one named road.

> I think someone mentioned Cambridge Uni was using OpenStreetMap for
> some of its maps,* so I'd be nervous about proposing anything radical
> right now. But is there anyone on this list who is a Cambridge mapper,
> or connected to the university's use of mapping? It's possible that
> some team decided to use the tag to mark every college building (etc),
> when really amenity=university is supposed to mark a university, not a
> piece of a university.
> 
> To do it "properly" it might need some neat relations to group these
> things. (Might be fun for someone who loves relations - various
> multi-site and hierarchical connections among the buildings scattered
> across town!) Alternatively there are tags in use such as
> building=university which might be good drop-in replacements...

I have been doing some things around the University of Southampton, I
have just made an attempt to bring the wiki page a bit more up to date
[1]. It should now cover some of the basics.

As for the tagging, for the University of Southampton, none of the
software that I have written (e.g. [2]) relies on amenity=university. I
am pretty much only relying on the URI's (uri tag) to link out of
OpenStreetMap. This allows for going in to OpenStreetMap and identifying
bits of information.

In summary, I am not sure what amenity=university means, or should mean,
however, it obviously plays a part in rendering OSM. I am unsure about
relations, I think it remains to be seen if they are fit for purpose in
this regard.

1: https://wiki.openstreetmap.org/wiki/University_of_Southampton
2: http://maps.southampton.ac.uk/




signature.asc
Description: OpenPGP digital signature
___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] too many universities in Cambridge

[Dan S]

Right, OK thanks. So let me try and answer without raising any
sub-controversies - that's why I was reluctant to answer "what would
you show for cambridge"!

If there was a single mortar-board for every geographically
self-contained UoC "site" on the map - that seems rather reasonable.
The minimalist way to achieve that would be for those sites to have
amenity=university tag, and for none of the buildings within them to
have that tag. (The humanitarian map would then look good...) I
personally would find that still a little bit curious but here I'm not
proposing to impose my ideal relation-tastic solution, since you've
raised some objections to that kind of thing.

For university buildings that are standalone, not part of a larger
"site" - well I guess if I had to design a map I wouldn't put any
mortar-board for them, though I might decide to give them a
mortar-board at the highest zoom level. (This might be achieved via
building=university perhaps. Though the question is about the
rendering not the tagging.)

Is this a meaningful answer to your rendering question? I hope so.

Best
Dan


2015-05-22 13:03 GMT+01:00 David Earl :
> Sorry, that wasn't intended to be provocative, it was a serious question.
> Irrespective of how it is tagged, how should one show a spread out
> institution on a map? If you do ARU with two mortar boards or some such
> should Cambridge be 10, one for each site, 41 including the colleges, or
> what? One could argue that it's the mapping you cited that's inadequate
> because it should collapse them into one when they are sufficiently close
> together to not be distinct (like ios does for photo locations on a map for
> example*), and that when zoomed in you *do* want them to be shown
> separately. In any case neither the current scheme nor a relation scheme
> preclude that, they are currently group-able by operator (which is a much
> more sustainable way of relating them IMO than relations).
>
> I asked about the building=university rendering because it would be a shame
> to lose the university buildings as distinct on the main map, and I have no
> control over fixing that. No doubt someone would catch up with it
> eventually.
>
> I would have to go back to the code to see what the exact implications of
> removing the amenity tags are, it's three years since I wrote it. I am
> almost certain that changing building=yes to building=university is
> harmless, but if I then have to rely on it, we have to be careful that
> university libraries aren't tagged building=library for example as the
> information gets lost.
>
> David
>
> * in similar vein one of the developments that's been requested for the
> university map is that when you get a search hit where the result blobs are
> overlapping they should be merged into one. This is very hard to do, so it
> will cost a lot.
>
>
> On Fri, 22 May 2015 at 12:40 Dan S  wrote:
>>
>> 2015-05-22 12:33 GMT+01:00 David Earl :
>> >> to render a map using a UK-wide rendering scheme that indicates
>> >> each university prominently
>> >
>> > What *would* you show for Cambridge (or other spread out ones) for this?
>> > Where actually *is* the University of Cambridge? What would you show
>> > that
>> > the current tagging doesn't already achieve?
>>
>> It doesn't matter what I would do. The UoC tagging is inconsistent
>> with the tagging for other universities, in a way that means no-one
>> can currently design a UK-wide map render that can handle universities
>> properly.
>>
>> I understand that you don't like this erupting under your feet, but
>> I'm afraid that's what happens in wiki-like systems. Please, please be
>> happy that I'm a considerate map editor who tries to discuss rather
>> than just to edit. We have absolutely no guarantees that a map editor
>> who loves consistency but doesn't love communication will not break
>> your schema at any moment!
>>
>> I'd be really grateful if you could comment on my suggestion about
>> modifying the building tags.
>>
>> Best
>> Dan
>>
>>
>> > On Fri, 22 May 2015 at 12:30 David Earl 
>> > wrote:
>> >>
>> >> Does the main OSM rendering understand building=university?
>> >>
>> >>
>> >> On Fri, 22 May 2015 at 12:27 Dan S  wrote:
>> >>>
>> >>> Hi David,
>> >>>
>> >>> Thanks for the detailed info. My main concern is in terms of the
>> >>> consequences for other data consumers. If someone tries to use OSM
>> >>> data for anything - such as:
>> >>>  (a) to plot the density of universities per county
>> >>>  (b) to render a map using a UK-wide rendering scheme that indicates
>> >>> each university prominently
>> >>> - the results will be utterly wrong. So we do need some consistency,
>> >>> at least at the country-wide level. (I'm pragmatic enough not to aim
>> >>> for global consistency ;)
>> >>>
>> >>> So I'm not offended, but I do care that our data is good for everyone.
>> >>> I suggest that we should make a change but I will not rush anything!
>> >>>
>> >>> I don't know what this "camp" is that didn't like buildin

Re: [Talk-GB] too many universities in Cambridge

[David Earl]

Sorry, that wasn't intended to be provocative, it was a serious question.
Irrespective of how it is tagged, how should one show a spread out
institution on a map? If you do ARU with two mortar boards or some such
should Cambridge be 10, one for each site, 41 including the colleges, or
what? One could argue that it's the mapping you cited that's inadequate
because it should collapse them into one when they are sufficiently close
together to not be distinct (like ios does for photo locations on a map for
example*), and that when zoomed in you *do* want them to be shown
separately. In any case neither the current scheme nor a relation scheme
preclude that, they are currently group-able by operator (which is a much
more sustainable way of relating them IMO than relations).

I asked about the building=university rendering because it would be a shame
to lose the university buildings as distinct on the main map, and I have no
control over fixing that. No doubt someone would catch up with it
eventually.

I would have to go back to the code to see what the exact implications of
removing the amenity tags are, it's three years since I wrote it. I am
almost certain that changing building=yes to building=university is
harmless, but if I then have to rely on it, we have to be careful that
university libraries aren't tagged building=library for example as the
information gets lost.

David

* in similar vein one of the developments that's been requested for the
university map is that when you get a search hit where the result blobs are
overlapping they should be merged into one. This is very hard to do, so it
will cost a lot.


On Fri, 22 May 2015 at 12:40 Dan S  wrote:

> 2015-05-22 12:33 GMT+01:00 David Earl :
> >> to render a map using a UK-wide rendering scheme that indicates
> >> each university prominently
> >
> > What *would* you show for Cambridge (or other spread out ones) for this?
> > Where actually *is* the University of Cambridge? What would you show that
> > the current tagging doesn't already achieve?
>
> It doesn't matter what I would do. The UoC tagging is inconsistent
> with the tagging for other universities, in a way that means no-one
> can currently design a UK-wide map render that can handle universities
> properly.
>
> I understand that you don't like this erupting under your feet, but
> I'm afraid that's what happens in wiki-like systems. Please, please be
> happy that I'm a considerate map editor who tries to discuss rather
> than just to edit. We have absolutely no guarantees that a map editor
> who loves consistency but doesn't love communication will not break
> your schema at any moment!
>
> I'd be really grateful if you could comment on my suggestion about
> modifying the building tags.
>
> Best
> Dan
>
>
> > On Fri, 22 May 2015 at 12:30 David Earl 
> wrote:
> >>
> >> Does the main OSM rendering understand building=university?
> >>
> >>
> >> On Fri, 22 May 2015 at 12:27 Dan S  wrote:
> >>>
> >>> Hi David,
> >>>
> >>> Thanks for the detailed info. My main concern is in terms of the
> >>> consequences for other data consumers. If someone tries to use OSM
> >>> data for anything - such as:
> >>>  (a) to plot the density of universities per county
> >>>  (b) to render a map using a UK-wide rendering scheme that indicates
> >>> each university prominently
> >>> - the results will be utterly wrong. So we do need some consistency,
> >>> at least at the country-wide level. (I'm pragmatic enough not to aim
> >>> for global consistency ;)
> >>>
> >>> So I'm not offended, but I do care that our data is good for everyone.
> >>> I suggest that we should make a change but I will not rush anything!
> >>>
> >>> I don't know what this "camp" is that didn't like building=university.
> >>> Was it an OSM camp (eg a discussion on the tagging email list)? Either
> >>> way, building=university is now used quite a lot worldwide
> >>> http://taginfo.openstreetmap.org/tags/building=university#map - I
> >>> think it's pretty uncontroversial as a building tag.
> >>>
> >>> So the question, I guess, is what "jobs" amenity=university is doing
> >>> in your scheme. Is it being used as a selector for extracting data? Is
> >>> it being used as a selector for rendering-rules? You've got your
> >>> operator=* tagging which looks good for selecting a data extract.
> >>>
> >>> If we made a two-step change such that all "building=yes,
> >>> amenity=university, operator=.*University of Cambridge.*" were first
> >>> modified to building=university, and then after a few months to remove
> >>> the amenity=university from buildings and leave it on
> >>> sites/universities/whatever-we-agree-it-should-be-on - would that work
> >>> for you? My quick overpass check suggests that would address about 800
> >>> of the 1200 objects.
> >>>
> >>> Best
> >>> Dan
> >>>
> >>>
> >>> 2015-05-22 11:54 GMT+01:00 David Earl :
> >>> > Hi Dan,
> >>> >
> >>> > Yes, Philip's right - I developed and continue to maintain the
> >>> > University
> >>> > map at 

Re: [Talk-GB] too many universities in Cambridge

[Lester Caine]

On 22/05/15 12:40, Dan S wrote:
> I'd be really grateful if you could comment on my suggestion about
> modifying the building tags.

Dan S ... The question still remains ... what are you trying to achieve?
Count the number of buildings making up a university, or something else?
In the distant past there were proposals for a much better hierarchy of
'places' where a place like 'University of Cambridge' would have a place
holder, and everything related to that place would be linked as such.
This probably pre-dates relations, but I STILL think that it has a
practical use today. When you select 'university, country' you see a
list of entities with a single entry per, but because it's not 'physical
map information' it's not acceptable to  some :(

-- 
Lester Caine - G8HFL
-
Contact - http://lsces.co.uk/wiki/?page=contact
L.S.Caine Electronic Services - http://lsces.co.uk
EnquirySolve - http://enquirysolve.com/
Model Engineers Digital Workshop - http://medw.co.uk
Rainbow Digital Media - http://rainbowdigitalmedia.co.uk

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] too many universities in Cambridge

[Paul Sladen]

On Fri, 22 May 2015, Dan S wrote:
> I'd be really grateful if you could comment on my suggestion about
> modifying the building tags.

David has commented (twice), both with follow-up questions.

-Paul



___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] too many universities in Cambridge

[Lester Caine]

On 22/05/15 12:26, David Earl wrote:
> If this discussion were happening at the start of the project four years
> ago. It wasn't as if the scheme wasn't public then. But it's been
> implemented now for several years, and to reorganise it is unhelpful and
> costly, with little benefit other than a sense of it "being right". (And
> in any group of 10 mappers, there seem to be 11 opinions as to what is
> right, concensus is very hard to achieve).

Only 11 :)

-- 
Lester Caine - G8HFL
-
Contact - http://lsces.co.uk/wiki/?page=contact
L.S.Caine Electronic Services - http://lsces.co.uk
EnquirySolve - http://enquirysolve.com/
Model Engineers Digital Workshop - http://medw.co.uk
Rainbow Digital Media - http://rainbowdigitalmedia.co.uk

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] too many universities in Cambridge

[Dan S]

2015-05-22 12:33 GMT+01:00 David Earl :
>> to render a map using a UK-wide rendering scheme that indicates
>> each university prominently
>
> What *would* you show for Cambridge (or other spread out ones) for this?
> Where actually *is* the University of Cambridge? What would you show that
> the current tagging doesn't already achieve?

It doesn't matter what I would do. The UoC tagging is inconsistent
with the tagging for other universities, in a way that means no-one
can currently design a UK-wide map render that can handle universities
properly.

I understand that you don't like this erupting under your feet, but
I'm afraid that's what happens in wiki-like systems. Please, please be
happy that I'm a considerate map editor who tries to discuss rather
than just to edit. We have absolutely no guarantees that a map editor
who loves consistency but doesn't love communication will not break
your schema at any moment!

I'd be really grateful if you could comment on my suggestion about
modifying the building tags.

Best
Dan


> On Fri, 22 May 2015 at 12:30 David Earl  wrote:
>>
>> Does the main OSM rendering understand building=university?
>>
>>
>> On Fri, 22 May 2015 at 12:27 Dan S  wrote:
>>>
>>> Hi David,
>>>
>>> Thanks for the detailed info. My main concern is in terms of the
>>> consequences for other data consumers. If someone tries to use OSM
>>> data for anything - such as:
>>>  (a) to plot the density of universities per county
>>>  (b) to render a map using a UK-wide rendering scheme that indicates
>>> each university prominently
>>> - the results will be utterly wrong. So we do need some consistency,
>>> at least at the country-wide level. (I'm pragmatic enough not to aim
>>> for global consistency ;)
>>>
>>> So I'm not offended, but I do care that our data is good for everyone.
>>> I suggest that we should make a change but I will not rush anything!
>>>
>>> I don't know what this "camp" is that didn't like building=university.
>>> Was it an OSM camp (eg a discussion on the tagging email list)? Either
>>> way, building=university is now used quite a lot worldwide
>>> http://taginfo.openstreetmap.org/tags/building=university#map - I
>>> think it's pretty uncontroversial as a building tag.
>>>
>>> So the question, I guess, is what "jobs" amenity=university is doing
>>> in your scheme. Is it being used as a selector for extracting data? Is
>>> it being used as a selector for rendering-rules? You've got your
>>> operator=* tagging which looks good for selecting a data extract.
>>>
>>> If we made a two-step change such that all "building=yes,
>>> amenity=university, operator=.*University of Cambridge.*" were first
>>> modified to building=university, and then after a few months to remove
>>> the amenity=university from buildings and leave it on
>>> sites/universities/whatever-we-agree-it-should-be-on - would that work
>>> for you? My quick overpass check suggests that would address about 800
>>> of the 1200 objects.
>>>
>>> Best
>>> Dan
>>>
>>>
>>> 2015-05-22 11:54 GMT+01:00 David Earl :
>>> > Hi Dan,
>>> >
>>> > Yes, Philip's right - I developed and continue to maintain the
>>> > University
>>> > map at http://map.cam.ac.uk (as well as doing all of the original
>>> > street
>>> > pattern mapping for Cambridge back in 2006). The University has put a
>>> > considerable investment and negotiated permission for college access
>>> > into
>>> > the map and contributed tens of thousands of pounds of survey data into
>>> > OSM
>>> > - it's not just "some of its maps", it's completely central to the
>>> > University map, not just a casual effort.
>>> >
>>> > The "schema" for tags that make the University map work is at
>>> > http://wiki.openstreetmap.org/wiki/Cambridge/University_of_Cambridge
>>> > (I've
>>> > just realised I haven't updated that page with a recent, unrelated new
>>> > bit,
>>> > I must do so).
>>> >
>>> > As it happens, I was also thinking about this the other day. The three
>>> > main
>>> > things are that is (a) consistent and (b) doesn't change under our feet
>>> > and
>>> > break the map, (c) it needs a way to distinguish these buildings from
>>> > others. It possibly wasn't the best decision to do it like this, though
>>> > I
>>> > still don't think it is a terrible way to do it. Relations would be
>>> > awful:
>>> > they are very hard to maintain accurately, and incidentally they are
>>> > hard to
>>> > work with in consumers because they come at the end of the data so you
>>> > have
>>> > to do multiple passes or keep lots of data in memory, and I think you'd
>>> > lose
>>> > most of the distinctive renderings off the OSM map, though since that's
>>> > such
>>> > an opaque process it's hard to know.
>>> >
>>> > building=university would work, but only if done in a controlled way so
>>> > that
>>> > we don't break the map while it's being done. But do you really want to
>>> > spend your days in front of a computer changing all the university tags
>>> > in
>>> > Cambridge though?! I can thi

Re: [Talk-GB] too many universities in Cambridge

[Dan S]

I don't think it renders it (though I thought it used to).

The humanitarian style renders the tag, and Cambridge looks mortar-board crazy:
http://www.openstreetmap.org/#map=15/52.2039/0.1182&layers=H

Dan


2015-05-22 12:30 GMT+01:00 David Earl :
> Does the main OSM rendering understand building=university?
>
>
> On Fri, 22 May 2015 at 12:27 Dan S  wrote:
>>
>> Hi David,
>>
>> Thanks for the detailed info. My main concern is in terms of the
>> consequences for other data consumers. If someone tries to use OSM
>> data for anything - such as:
>>  (a) to plot the density of universities per county
>>  (b) to render a map using a UK-wide rendering scheme that indicates
>> each university prominently
>> - the results will be utterly wrong. So we do need some consistency,
>> at least at the country-wide level. (I'm pragmatic enough not to aim
>> for global consistency ;)
>>
>> So I'm not offended, but I do care that our data is good for everyone.
>> I suggest that we should make a change but I will not rush anything!
>>
>> I don't know what this "camp" is that didn't like building=university.
>> Was it an OSM camp (eg a discussion on the tagging email list)? Either
>> way, building=university is now used quite a lot worldwide
>> http://taginfo.openstreetmap.org/tags/building=university#map - I
>> think it's pretty uncontroversial as a building tag.
>>
>> So the question, I guess, is what "jobs" amenity=university is doing
>> in your scheme. Is it being used as a selector for extracting data? Is
>> it being used as a selector for rendering-rules? You've got your
>> operator=* tagging which looks good for selecting a data extract.
>>
>> If we made a two-step change such that all "building=yes,
>> amenity=university, operator=.*University of Cambridge.*" were first
>> modified to building=university, and then after a few months to remove
>> the amenity=university from buildings and leave it on
>> sites/universities/whatever-we-agree-it-should-be-on - would that work
>> for you? My quick overpass check suggests that would address about 800
>> of the 1200 objects.
>>
>> Best
>> Dan
>>
>>
>> 2015-05-22 11:54 GMT+01:00 David Earl :
>> > Hi Dan,
>> >
>> > Yes, Philip's right - I developed and continue to maintain the
>> > University
>> > map at http://map.cam.ac.uk (as well as doing all of the original street
>> > pattern mapping for Cambridge back in 2006). The University has put a
>> > considerable investment and negotiated permission for college access
>> > into
>> > the map and contributed tens of thousands of pounds of survey data into
>> > OSM
>> > - it's not just "some of its maps", it's completely central to the
>> > University map, not just a casual effort.
>> >
>> > The "schema" for tags that make the University map work is at
>> > http://wiki.openstreetmap.org/wiki/Cambridge/University_of_Cambridge
>> > (I've
>> > just realised I haven't updated that page with a recent, unrelated new
>> > bit,
>> > I must do so).
>> >
>> > As it happens, I was also thinking about this the other day. The three
>> > main
>> > things are that is (a) consistent and (b) doesn't change under our feet
>> > and
>> > break the map, (c) it needs a way to distinguish these buildings from
>> > others. It possibly wasn't the best decision to do it like this, though
>> > I
>> > still don't think it is a terrible way to do it. Relations would be
>> > awful:
>> > they are very hard to maintain accurately, and incidentally they are
>> > hard to
>> > work with in consumers because they come at the end of the data so you
>> > have
>> > to do multiple passes or keep lots of data in memory, and I think you'd
>> > lose
>> > most of the distinctive renderings off the OSM map, though since that's
>> > such
>> > an opaque process it's hard to know.
>> >
>> > building=university would work, but only if done in a controlled way so
>> > that
>> > we don't break the map while it's being done. But do you really want to
>> > spend your days in front of a computer changing all the university tags
>> > in
>> > Cambridge though?! I can think of more productive and helpful things to
>> > do.
>> > I did consider building=university, but like all things OSM, there was a
>> > camp that only wanted building=yes, and that is what the Map_features
>> > page
>> > then decreed (it has more now, but university isn't among them). The
>> > more
>> > critical tags from my point of view are the operator ones.
>> >
>> > This raises some other points though...
>> >
>> > 1. What about the sites and the colleges? These are also tagged
>> > University,
>> > and there isn't an obvious alternative that won't mean the ordinary OSM
>> > maps
>> > don't show them. Fundamentally, is a "part of a university" a
>> > university? I
>> > think it's helpful to do it like that. Did you know the University of
>> > Nottingham has a branch in China - would it really be helpful to link
>> > these
>> > with relations spanning the world? I think there's cases both ways.
>> >
>> > 2. What is a Un

Re: [Talk-GB] too many universities in Cambridge

[David Earl]

> to render a map using a UK-wide rendering scheme that indicates
> each university prominently

What *would* you show for Cambridge (or other spread out ones) for this?
Where actually *is* the University of Cambridge? What would you show that
the current tagging doesn't already achieve?

On Fri, 22 May 2015 at 12:30 David Earl  wrote:

> Does the main OSM rendering understand building=university?
>
>
> On Fri, 22 May 2015 at 12:27 Dan S  wrote:
>
>> Hi David,
>>
>> Thanks for the detailed info. My main concern is in terms of the
>> consequences for other data consumers. If someone tries to use OSM
>> data for anything - such as:
>>  (a) to plot the density of universities per county
>>  (b) to render a map using a UK-wide rendering scheme that indicates
>> each university prominently
>> - the results will be utterly wrong. So we do need some consistency,
>> at least at the country-wide level. (I'm pragmatic enough not to aim
>> for global consistency ;)
>>
>> So I'm not offended, but I do care that our data is good for everyone.
>> I suggest that we should make a change but I will not rush anything!
>>
>> I don't know what this "camp" is that didn't like building=university.
>> Was it an OSM camp (eg a discussion on the tagging email list)? Either
>> way, building=university is now used quite a lot worldwide
>> http://taginfo.openstreetmap.org/tags/building=university#map - I
>> think it's pretty uncontroversial as a building tag.
>>
>> So the question, I guess, is what "jobs" amenity=university is doing
>> in your scheme. Is it being used as a selector for extracting data? Is
>> it being used as a selector for rendering-rules? You've got your
>> operator=* tagging which looks good for selecting a data extract.
>>
>> If we made a two-step change such that all "building=yes,
>> amenity=university, operator=.*University of Cambridge.*" were first
>> modified to building=university, and then after a few months to remove
>> the amenity=university from buildings and leave it on
>> sites/universities/whatever-we-agree-it-should-be-on - would that work
>> for you? My quick overpass check suggests that would address about 800
>> of the 1200 objects.
>>
>> Best
>> Dan
>>
>>
>> 2015-05-22 11:54 GMT+01:00 David Earl :
>> > Hi Dan,
>> >
>> > Yes, Philip's right - I developed and continue to maintain the
>> University
>> > map at http://map.cam.ac.uk (as well as doing all of the original
>> street
>> > pattern mapping for Cambridge back in 2006). The University has put a
>> > considerable investment and negotiated permission for college access
>> into
>> > the map and contributed tens of thousands of pounds of survey data into
>> OSM
>> > - it's not just "some of its maps", it's completely central to the
>> > University map, not just a casual effort.
>> >
>> > The "schema" for tags that make the University map work is at
>> > http://wiki.openstreetmap.org/wiki/Cambridge/University_of_Cambridge
>> (I've
>> > just realised I haven't updated that page with a recent, unrelated new
>> bit,
>> > I must do so).
>> >
>> > As it happens, I was also thinking about this the other day. The three
>> main
>> > things are that is (a) consistent and (b) doesn't change under our feet
>> and
>> > break the map, (c) it needs a way to distinguish these buildings from
>> > others. It possibly wasn't the best decision to do it like this, though
>> I
>> > still don't think it is a terrible way to do it. Relations would be
>> awful:
>> > they are very hard to maintain accurately, and incidentally they are
>> hard to
>> > work with in consumers because they come at the end of the data so you
>> have
>> > to do multiple passes or keep lots of data in memory, and I think you'd
>> lose
>> > most of the distinctive renderings off the OSM map, though since that's
>> such
>> > an opaque process it's hard to know.
>> >
>> > building=university would work, but only if done in a controlled way so
>> that
>> > we don't break the map while it's being done. But do you really want to
>> > spend your days in front of a computer changing all the university tags
>> in
>> > Cambridge though?! I can think of more productive and helpful things to
>> do.
>> > I did consider building=university, but like all things OSM, there was a
>> > camp that only wanted building=yes, and that is what the Map_features
>> page
>> > then decreed (it has more now, but university isn't among them). The
>> more
>> > critical tags from my point of view are the operator ones.
>> >
>> > This raises some other points though...
>> >
>> > 1. What about the sites and the colleges? These are also tagged
>> University,
>> > and there isn't an obvious alternative that won't mean the ordinary OSM
>> maps
>> > don't show them. Fundamentally, is a "part of a university" a
>> university? I
>> > think it's helpful to do it like that. Did you know the University of
>> > Nottingham has a branch in China - would it really be helpful to link
>> these
>> > with relations spanning the world? I think t

Re: [Talk-GB] too many universities in Cambridge

[David Earl]

Does the main OSM rendering understand building=university?


On Fri, 22 May 2015 at 12:27 Dan S  wrote:

> Hi David,
>
> Thanks for the detailed info. My main concern is in terms of the
> consequences for other data consumers. If someone tries to use OSM
> data for anything - such as:
>  (a) to plot the density of universities per county
>  (b) to render a map using a UK-wide rendering scheme that indicates
> each university prominently
> - the results will be utterly wrong. So we do need some consistency,
> at least at the country-wide level. (I'm pragmatic enough not to aim
> for global consistency ;)
>
> So I'm not offended, but I do care that our data is good for everyone.
> I suggest that we should make a change but I will not rush anything!
>
> I don't know what this "camp" is that didn't like building=university.
> Was it an OSM camp (eg a discussion on the tagging email list)? Either
> way, building=university is now used quite a lot worldwide
> http://taginfo.openstreetmap.org/tags/building=university#map - I
> think it's pretty uncontroversial as a building tag.
>
> So the question, I guess, is what "jobs" amenity=university is doing
> in your scheme. Is it being used as a selector for extracting data? Is
> it being used as a selector for rendering-rules? You've got your
> operator=* tagging which looks good for selecting a data extract.
>
> If we made a two-step change such that all "building=yes,
> amenity=university, operator=.*University of Cambridge.*" were first
> modified to building=university, and then after a few months to remove
> the amenity=university from buildings and leave it on
> sites/universities/whatever-we-agree-it-should-be-on - would that work
> for you? My quick overpass check suggests that would address about 800
> of the 1200 objects.
>
> Best
> Dan
>
>
> 2015-05-22 11:54 GMT+01:00 David Earl :
> > Hi Dan,
> >
> > Yes, Philip's right - I developed and continue to maintain the University
> > map at http://map.cam.ac.uk (as well as doing all of the original street
> > pattern mapping for Cambridge back in 2006). The University has put a
> > considerable investment and negotiated permission for college access into
> > the map and contributed tens of thousands of pounds of survey data into
> OSM
> > - it's not just "some of its maps", it's completely central to the
> > University map, not just a casual effort.
> >
> > The "schema" for tags that make the University map work is at
> > http://wiki.openstreetmap.org/wiki/Cambridge/University_of_Cambridge
> (I've
> > just realised I haven't updated that page with a recent, unrelated new
> bit,
> > I must do so).
> >
> > As it happens, I was also thinking about this the other day. The three
> main
> > things are that is (a) consistent and (b) doesn't change under our feet
> and
> > break the map, (c) it needs a way to distinguish these buildings from
> > others. It possibly wasn't the best decision to do it like this, though I
> > still don't think it is a terrible way to do it. Relations would be
> awful:
> > they are very hard to maintain accurately, and incidentally they are
> hard to
> > work with in consumers because they come at the end of the data so you
> have
> > to do multiple passes or keep lots of data in memory, and I think you'd
> lose
> > most of the distinctive renderings off the OSM map, though since that's
> such
> > an opaque process it's hard to know.
> >
> > building=university would work, but only if done in a controlled way so
> that
> > we don't break the map while it's being done. But do you really want to
> > spend your days in front of a computer changing all the university tags
> in
> > Cambridge though?! I can think of more productive and helpful things to
> do.
> > I did consider building=university, but like all things OSM, there was a
> > camp that only wanted building=yes, and that is what the Map_features
> page
> > then decreed (it has more now, but university isn't among them). The more
> > critical tags from my point of view are the operator ones.
> >
> > This raises some other points though...
> >
> > 1. What about the sites and the colleges? These are also tagged
> University,
> > and there isn't an obvious alternative that won't mean the ordinary OSM
> maps
> > don't show them. Fundamentally, is a "part of a university" a
> university? I
> > think it's helpful to do it like that. Did you know the University of
> > Nottingham has a branch in China - would it really be helpful to link
> these
> > with relations spanning the world? I think there's cases both ways.
> >
> > 2. What is a University anyway? Almost no university is in one physical
> > area. Even campus universities like UEA have outlying premises (in UEA's
> > case in London too). Do you really not want the campus area to be tagged
> > university just because it isn't the whole thing? You said Anglia Ruskin
> was
> > one of the two universities in Cambridge - no it isn't, it's HALF a
> > university, the rest is in Chelmsford. I don

Re: [Talk-GB] too many universities in Cambridge

[Dan S]

Hi David,

Thanks for the detailed info. My main concern is in terms of the
consequences for other data consumers. If someone tries to use OSM
data for anything - such as:
 (a) to plot the density of universities per county
 (b) to render a map using a UK-wide rendering scheme that indicates
each university prominently
- the results will be utterly wrong. So we do need some consistency,
at least at the country-wide level. (I'm pragmatic enough not to aim
for global consistency ;)

So I'm not offended, but I do care that our data is good for everyone.
I suggest that we should make a change but I will not rush anything!

I don't know what this "camp" is that didn't like building=university.
Was it an OSM camp (eg a discussion on the tagging email list)? Either
way, building=university is now used quite a lot worldwide
http://taginfo.openstreetmap.org/tags/building=university#map - I
think it's pretty uncontroversial as a building tag.

So the question, I guess, is what "jobs" amenity=university is doing
in your scheme. Is it being used as a selector for extracting data? Is
it being used as a selector for rendering-rules? You've got your
operator=* tagging which looks good for selecting a data extract.

If we made a two-step change such that all "building=yes,
amenity=university, operator=.*University of Cambridge.*" were first
modified to building=university, and then after a few months to remove
the amenity=university from buildings and leave it on
sites/universities/whatever-we-agree-it-should-be-on - would that work
for you? My quick overpass check suggests that would address about 800
of the 1200 objects.

Best
Dan


2015-05-22 11:54 GMT+01:00 David Earl :
> Hi Dan,
>
> Yes, Philip's right - I developed and continue to maintain the University
> map at http://map.cam.ac.uk (as well as doing all of the original street
> pattern mapping for Cambridge back in 2006). The University has put a
> considerable investment and negotiated permission for college access into
> the map and contributed tens of thousands of pounds of survey data into OSM
> - it's not just "some of its maps", it's completely central to the
> University map, not just a casual effort.
>
> The "schema" for tags that make the University map work is at
> http://wiki.openstreetmap.org/wiki/Cambridge/University_of_Cambridge (I've
> just realised I haven't updated that page with a recent, unrelated new bit,
> I must do so).
>
> As it happens, I was also thinking about this the other day. The three main
> things are that is (a) consistent and (b) doesn't change under our feet and
> break the map, (c) it needs a way to distinguish these buildings from
> others. It possibly wasn't the best decision to do it like this, though I
> still don't think it is a terrible way to do it. Relations would be awful:
> they are very hard to maintain accurately, and incidentally they are hard to
> work with in consumers because they come at the end of the data so you have
> to do multiple passes or keep lots of data in memory, and I think you'd lose
> most of the distinctive renderings off the OSM map, though since that's such
> an opaque process it's hard to know.
>
> building=university would work, but only if done in a controlled way so that
> we don't break the map while it's being done. But do you really want to
> spend your days in front of a computer changing all the university tags in
> Cambridge though?! I can think of more productive and helpful things to do.
> I did consider building=university, but like all things OSM, there was a
> camp that only wanted building=yes, and that is what the Map_features page
> then decreed (it has more now, but university isn't among them). The more
> critical tags from my point of view are the operator ones.
>
> This raises some other points though...
>
> 1. What about the sites and the colleges? These are also tagged University,
> and there isn't an obvious alternative that won't mean the ordinary OSM maps
> don't show them. Fundamentally, is a "part of a university" a university? I
> think it's helpful to do it like that. Did you know the University of
> Nottingham has a branch in China - would it really be helpful to link these
> with relations spanning the world? I think there's cases both ways.
>
> 2. What is a University anyway? Almost no university is in one physical
> area. Even campus universities like UEA have outlying premises (in UEA's
> case in London too). Do you really not want the campus area to be tagged
> university just because it isn't the whole thing? You said Anglia Ruskin was
> one of the two universities in Cambridge - no it isn't, it's HALF a
> university, the rest is in Chelmsford. I don't think it would do any harm
> and would be helpful to group them with relations, if that were maintainable
> sustainably, but not at the expense of losing the tags from the outline
> itself. And the building thing only extends this further. Is a University a
> geographical thing at all? It's an institution, whi

Re: [Talk-GB] too many universities in Cambridge

[David Earl]

If this discussion were happening at the start of the project four years
ago. It wasn't as if the scheme wasn't public then. But it's been
implemented now for several years, and to reorganise it is unhelpful and
costly, with little benefit other than a sense of it "being right". (And in
any group of 10 mappers, there seem to be 11 opinions as to what is right,
concensus is very hard to achieve).

On Fri, 22 May 2015 at 12:22 Richard Mann 
wrote:

> A quick scan of Oxford shows the colleges (and a few multi-building areas
> such as the Science Area) as amenity=university, with buildings within
> colleges and odd departments as building=university. So we have a lot of
> universities too.
>
> Other big difference is that we haven't generally added "(University of
> Oxford)" to the end of all the college names...
>
> I'd tend to go for amenity=university for a contiguous site with a single
> name, with the occasional split site (eg on two sides of a public road) as
> a multi-polygon. Then I'd add a *tag* to show that the site was part of a
> collection making up the University (probably operator, though that feels
> wrong, since the colleges are independent entities). It's *not* a candidate
> for a relation because there are no geographical relationships between the
> components.
>
> Richard
>
> On Fri, May 22, 2015 at 11:54 AM, David Earl 
> wrote:
>
>> Hi Dan,
>>
>> Yes, Philip's right - I developed and continue to maintain the University
>> map at http://map.cam.ac.uk (as well as doing all of the original street
>> pattern mapping for Cambridge back in 2006). The University has put a
>> considerable investment and negotiated permission for college access into
>> the map and contributed tens of thousands of pounds of survey data into OSM
>> - it's not just "some of its maps", it's completely central to the
>> University map, not just a casual effort.
>>
>> The "schema" for tags that make the University map work is at
>> http://wiki.openstreetmap.org/wiki/Cambridge/University_of_Cambridge
>> (I've just realised I haven't updated that page with a recent, unrelated
>> new bit, I must do so).
>>
>> As it happens, I was also thinking about this the other day. The three
>> main things are that is (a) consistent and (b) doesn't change under our
>> feet and break the map, (c) it needs a way to distinguish these buildings
>> from others. It possibly wasn't the best decision to do it like this,
>> though I still don't think it is a terrible way to do it. Relations would
>> be awful: they are very hard to maintain accurately, and incidentally they
>> are hard to work with in consumers because they come at the end of the data
>> so you have to do multiple passes or keep lots of data in memory, and I
>> think you'd lose most of the distinctive renderings off the OSM map, though
>> since that's such an opaque process it's hard to know.
>>
>> building=university would work, but only if done in a controlled way so
>> that we don't break the map while it's being done. But do you really want
>> to spend your days in front of a computer changing all the university tags
>> in Cambridge though?! I can think of more productive and helpful things to
>> do. I did consider building=university, but like all things OSM, there was
>> a camp that only wanted building=yes, and that is what the Map_features
>> page then decreed (it has more now, but university isn't among them). The
>> more critical tags from my point of view are the operator ones.
>>
>> This raises some other points though...
>>
>> 1. What about the sites and the colleges? These are also tagged
>> University, and there isn't an obvious alternative that won't mean the
>> ordinary OSM maps don't show them. Fundamentally, is a "part of a
>> university" a university? I think it's helpful to do it like that. Did you
>> know the University of Nottingham has a branch in China - would it really
>> be helpful to link these with relations spanning the world? I think there's
>> cases both ways.
>>
>> 2. What is a University anyway? Almost no university is in one physical
>> area. Even campus universities like UEA have outlying premises (in UEA's
>> case in London too). Do you really not want the campus area to be tagged
>> university just because it isn't the whole thing? You said Anglia Ruskin
>> was one of the two universities in Cambridge - no it isn't, it's HALF a
>> university, the rest is in Chelmsford. I don't think it would do any harm
>> and would be helpful to group them with relations, if that were
>> maintainable sustainably, but not at the expense of losing the tags from
>> the outline itself. And the building thing only extends this further. Is
>> a University a geographical thing at all? It's an institution, which may
>> have some buildings but really it's a concept not a physical object -
>> ultimately everything on the map is just a part, not the whole.
>>
>> 4. Constantly changing tags creates a moving target that is extremely
>> hard to maintain for data consumer

Re: [Talk-GB] too many universities in Cambridge

[Richard Mann]

A quick scan of Oxford shows the colleges (and a few multi-building areas
such as the Science Area) as amenity=university, with buildings within
colleges and odd departments as building=university. So we have a lot of
universities too.

Other big difference is that we haven't generally added "(University of
Oxford)" to the end of all the college names...

I'd tend to go for amenity=university for a contiguous site with a single
name, with the occasional split site (eg on two sides of a public road) as
a multi-polygon. Then I'd add a *tag* to show that the site was part of a
collection making up the University (probably operator, though that feels
wrong, since the colleges are independent entities). It's *not* a candidate
for a relation because there are no geographical relationships between the
components.

Richard

On Fri, May 22, 2015 at 11:54 AM, David Earl 
wrote:

> Hi Dan,
>
> Yes, Philip's right - I developed and continue to maintain the University
> map at http://map.cam.ac.uk (as well as doing all of the original street
> pattern mapping for Cambridge back in 2006). The University has put a
> considerable investment and negotiated permission for college access into
> the map and contributed tens of thousands of pounds of survey data into OSM
> - it's not just "some of its maps", it's completely central to the
> University map, not just a casual effort.
>
> The "schema" for tags that make the University map work is at
> http://wiki.openstreetmap.org/wiki/Cambridge/University_of_Cambridge
> (I've just realised I haven't updated that page with a recent, unrelated
> new bit, I must do so).
>
> As it happens, I was also thinking about this the other day. The three
> main things are that is (a) consistent and (b) doesn't change under our
> feet and break the map, (c) it needs a way to distinguish these buildings
> from others. It possibly wasn't the best decision to do it like this,
> though I still don't think it is a terrible way to do it. Relations would
> be awful: they are very hard to maintain accurately, and incidentally they
> are hard to work with in consumers because they come at the end of the data
> so you have to do multiple passes or keep lots of data in memory, and I
> think you'd lose most of the distinctive renderings off the OSM map, though
> since that's such an opaque process it's hard to know.
>
> building=university would work, but only if done in a controlled way so
> that we don't break the map while it's being done. But do you really want
> to spend your days in front of a computer changing all the university tags
> in Cambridge though?! I can think of more productive and helpful things to
> do. I did consider building=university, but like all things OSM, there was
> a camp that only wanted building=yes, and that is what the Map_features
> page then decreed (it has more now, but university isn't among them). The
> more critical tags from my point of view are the operator ones.
>
> This raises some other points though...
>
> 1. What about the sites and the colleges? These are also tagged
> University, and there isn't an obvious alternative that won't mean the
> ordinary OSM maps don't show them. Fundamentally, is a "part of a
> university" a university? I think it's helpful to do it like that. Did you
> know the University of Nottingham has a branch in China - would it really
> be helpful to link these with relations spanning the world? I think there's
> cases both ways.
>
> 2. What is a University anyway? Almost no university is in one physical
> area. Even campus universities like UEA have outlying premises (in UEA's
> case in London too). Do you really not want the campus area to be tagged
> university just because it isn't the whole thing? You said Anglia Ruskin
> was one of the two universities in Cambridge - no it isn't, it's HALF a
> university, the rest is in Chelmsford. I don't think it would do any harm
> and would be helpful to group them with relations, if that were
> maintainable sustainably, but not at the expense of losing the tags from
> the outline itself. And the building thing only extends this further. Is
> a University a geographical thing at all? It's an institution, which may
> have some buildings but really it's a concept not a physical object -
> ultimately everything on the map is just a part, not the whole.
>
> 4. Constantly changing tags creates a moving target that is extremely hard
> to maintain for data consumers, and is a major off-putting factor in using
> OSM, especially if you can't manage the process because things just change
> under your feet. For example, there is a thread on talk discussing
> completely changing the amenities altogether, without regard for people who
> want to use this stuff in the real world. My view is that tags are merely
> tokens and too much is read into the words. They are part of the API and
> the fact you can change them because you prefer some other structure
> doesn't mean you should. The flexibility means we can 

Re: [Talk-GB] too many universities in Cambridge

[Lester Caine]

On 22/05/15 12:00, David Earl wrote:
> The "schema" for tags that make the University map work is at
> http://wiki.openstreetmap.org/wiki/Cambridge/University_of_Cambridge
> (I've just realised I haven't updated that page with a recent,
> unrelated new bit, I must do so).
> 
> Oh, I did, I'd forgotten! It's this bit at the end I
> meant: 
> http://wiki.openstreetmap.org/wiki/Cambridge/University_of_Cambridge#Non-university_references
> . It's what makes the red buildings on the University map (like ARU).

That makes sense on a number of disjointed facilities. I've not looked
at Warwick University recently, but it has acquired satellite locations
over the years. But Hospitals are other entities that would benefit from
the same treatment, and the =reference element does seem to be the way
many of these sprawling installations now help to show visitors just
where to park for a particular department. A couple of my regular haunts
have just completed a re-signing exercise to help identification of
departments on a number of levels.

-- 
Lester Caine - G8HFL
-
Contact - http://lsces.co.uk/wiki/?page=contact
L.S.Caine Electronic Services - http://lsces.co.uk
EnquirySolve - http://enquirysolve.com/
Model Engineers Digital Workshop - http://medw.co.uk
Rainbow Digital Media - http://rainbowdigitalmedia.co.uk

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] too many universities in Cambridge

[David Earl]

On Fri, 22 May 2015 at 11:54 David Earl  wrote:

> The "schema" for tags that make the University map work is at
> http://wiki.openstreetmap.org/wiki/Cambridge/University_of_Cambridge
> (I've just realised I haven't updated that page with a recent, unrelated
> new bit, I must do so).
>

Oh, I did, I'd forgotten! It's this bit at the end I meant:
http://wiki.openstreetmap.org/wiki/Cambridge/University_of_Cambridge#Non-university_references
. It's what makes the red buildings on the University map (like ARU).
___
Talk-GB mailing list
Talk-GB@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-gb

Re: [Talk-GB] too many universities in Cambridge

[David Earl]

Hi Dan,

Yes, Philip's right - I developed and continue to maintain the University
map at http://map.cam.ac.uk (as well as doing all of the original street
pattern mapping for Cambridge back in 2006). The University has put a
considerable investment and negotiated permission for college access into
the map and contributed tens of thousands of pounds of survey data into OSM
- it's not just "some of its maps", it's completely central to the
University map, not just a casual effort.

The "schema" for tags that make the University map work is at
http://wiki.openstreetmap.org/wiki/Cambridge/University_of_Cambridge (I've
just realised I haven't updated that page with a recent, unrelated new bit,
I must do so).

As it happens, I was also thinking about this the other day. The three main
things are that is (a) consistent and (b) doesn't change under our feet and
break the map, (c) it needs a way to distinguish these buildings from
others. It possibly wasn't the best decision to do it like this, though I
still don't think it is a terrible way to do it. Relations would be awful:
they are very hard to maintain accurately, and incidentally they are hard
to work with in consumers because they come at the end of the data so you
have to do multiple passes or keep lots of data in memory, and I think
you'd lose most of the distinctive renderings off the OSM map, though since
that's such an opaque process it's hard to know.

building=university would work, but only if done in a controlled way so
that we don't break the map while it's being done. But do you really want
to spend your days in front of a computer changing all the university tags
in Cambridge though?! I can think of more productive and helpful things to
do. I did consider building=university, but like all things OSM, there was
a camp that only wanted building=yes, and that is what the Map_features
page then decreed (it has more now, but university isn't among them). The
more critical tags from my point of view are the operator ones.

This raises some other points though...

1. What about the sites and the colleges? These are also tagged University,
and there isn't an obvious alternative that won't mean the ordinary OSM
maps don't show them. Fundamentally, is a "part of a university" a
university? I think it's helpful to do it like that. Did you know the
University of Nottingham has a branch in China - would it really be helpful
to link these with relations spanning the world? I think there's cases both
ways.

2. What is a University anyway? Almost no university is in one physical
area. Even campus universities like UEA have outlying premises (in UEA's
case in London too). Do you really not want the campus area to be tagged
university just because it isn't the whole thing? You said Anglia Ruskin
was one of the two universities in Cambridge - no it isn't, it's HALF a
university, the rest is in Chelmsford. I don't think it would do any harm
and would be helpful to group them with relations, if that were
maintainable sustainably, but not at the expense of losing the tags from
the outline itself. And the building thing only extends this further. Is a
University a geographical thing at all? It's an institution, which may have
some buildings but really it's a concept not a physical object - ultimately
everything on the map is just a part, not the whole.

4. Constantly changing tags creates a moving target that is extremely hard
to maintain for data consumers, and is a major off-putting factor in using
OSM, especially if you can't manage the process because things just change
under your feet. For example, there is a thread on talk discussing
completely changing the amenities altogether, without regard for people who
want to use this stuff in the real world. My view is that tags are merely
tokens and too much is read into the words. They are part of the API and
the fact you can change them because you prefer some other structure
doesn't mean you should. The flexibility means we can introduce new things
easily, but constant change is hard to cope with. The costs are borne
elsewhere, and what really does it buy us?

So, I think it's OK the way it is. If it offends you unbearably,
building=university wouldn't be too hard to cope with, but please, please
don't just do it, let me change the University software first, otherwise
the map will be broken on next update (which are frequent) and they will be
very annoyed. As I said, this is effectively part of the API, even though
it may not feel like it, and constitutes a non-upward compatible change. If
you do want to do it, please do it all, not in bits, and bear in mind this
has a direct financial cost to me as a freelancer supporting the University
map, and that the University has been a big benefactor for OSM, even though
they get the rest of the map back in return, so you really don't want to
give them a slap in the face for doing so.

David


On Thu, 21 May 2015 at 23:13 Phillip Barnett 
wrote:

> I'm a Cambridge mapper, but